Hunting unicorns with Network analysis
summary
Constructs a monthly time‑varying global startup network from CrunchBase records (1990–2015, ~36,000 people, ~284,000 organizations, 117 countries) by linking companies when employees move between them, implemented with NetworkX. Uses PageRank centrality to rescore and rank pre‑seed startups (defined <2 years old and pre‑seed) and measures outcomes (acquisition, IPO, or acquiring another company) within seven years. Results double early‑stage fund performance for a top‑20 shortlist (≈30% success vs ~15% baseline), so the pipeline is concrete, reproducible, and shows measurable predictive value.
AI-written summary · about
chapters (16)
- 0:00Introduction
- 1:03Speaker background
- 4:15Talk roadmap
- 5:18Startup success challenges
- 7:26Success factors hypothesis
- 8:28Knowledge transfer network
- 10:59Startup network visualization
- 11:20Network centrality explained
- 13:08Ranking without data leakage
- 15:12Startup success metrics
- 16:38Ranking performance evaluation
- 18:27Results versus VC funds
- 20:16Early unicorn detections
- 21:19Practical deliverables for venture capitalists
- 23:27Take-home message
- 25:38Q&A
related talks
23mOn the startup line
24mHands-On Network Science
28mThe Future of VC - How the Industry Gets Disrupted
23mUsing Software, AI, DS to Source Growth Opportunities
36mHappiness inside a job: a social network analysis
28mNetworks meet Finance in Python
40mDiscovering Emerging Open Source Communities Through...
01h 38mNetwork Science for Fun and Profit